Migrate String.prototype.to{Upper,Lower}Case functions from JS to CPP builtins.

Move ICU case conversion utility functions to a common location.

BUG=v8:5751
CQ_INCLUDE_TRYBOTS=master.tryserver.v8:v8_linux_noi18n_rel_ng

Review-Url: https://codereview.chromium.org/2728763006
Cr-Commit-Position: refs/heads/master@{#44050}

BUILD.gn

@@ -1115,6 +1115,7 @@ v8_source_set("v8_base") {
     "src/builtins/builtins-global.cc",
     "src/builtins/builtins-internal.cc",
     "src/builtins/builtins-interpreter.cc",
+    "src/builtins/builtins-intl.cc",
     "src/builtins/builtins-json.cc",
     "src/builtins/builtins-math.cc",
     "src/builtins/builtins-number.cc",

src/bootstrapper.cc

@@ -3890,21 +3890,22 @@ void Genesis::InitializeGlobal_icu_case_mapping() {
   Handle<JSObject> string_prototype(
       JSObject::cast(native_context()->string_function()->prototype()));
 
-  Handle<JSFunction> to_lower_case = Handle<JSFunction>::cast(
-      JSReceiver::GetProperty(
-          exports_container,
-          factory()->InternalizeUtf8String("ToLowerCaseI18N"))
-          .ToHandleChecked());
-  SetFunction(string_prototype, to_lower_case,
-              factory()->InternalizeUtf8String("toLowerCase"));
-
-  Handle<JSFunction> to_upper_case = Handle<JSFunction>::cast(
-      JSReceiver::GetProperty(
-          exports_container,
-          factory()->InternalizeUtf8String("ToUpperCaseI18N"))
-          .ToHandleChecked());
-  SetFunction(string_prototype, to_upper_case,
-              factory()->InternalizeUtf8String("toUpperCase"));
+  {
+    Handle<String> name = factory()->InternalizeUtf8String("toLowerCase");
+    SetFunction(string_prototype,
+                SimpleCreateFunction(isolate(), name,
+                                     Builtins::kStringPrototypeToLowerCaseI18N,
+                                     0, false),
+                name);
+  }
+  {
+    Handle<String> name = factory()->InternalizeUtf8String("toUpperCase");
+    SetFunction(string_prototype,
+                SimpleCreateFunction(isolate(), name,
+                                     Builtins::kStringPrototypeToUpperCaseI18N,
+                                     0, false),
+                name);
+  }
 
   Handle<JSFunction> to_locale_lower_case = Handle<JSFunction>::cast(
       JSReceiver::GetProperty(

src/builtins/builtins-intl.cc

+// Copyright 2017 the V8 project authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifdef V8_I18N_SUPPORT
+
+#include "src/builtins/builtins-utils.h"
+#include "src/builtins/builtins.h"
+#include "src/i18n.h"
+#include "src/objects-inl.h"
+
+namespace v8 {
+namespace internal {
+
+BUILTIN(StringPrototypeToLowerCaseI18N) {
+  HandleScope scope(isolate);
+  TO_THIS_STRING(string, "String.prototype.toLowerCase");
+  string = String::Flatten(string);
+  return ConvertCase(string, false, isolate);
+}
+
+BUILTIN(StringPrototypeToUpperCaseI18N) {
+  HandleScope scope(isolate);
+  TO_THIS_STRING(string, "String.prototype.toUpperCase");
+  string = String::Flatten(string);
+  return ConvertCase(string, true, isolate);
+}
+
+}  // namespace internal
+}  // namespace v8
+
+#endif  // V8_I18N_SUPPORT

src/builtins/builtins.h

@@ -51,7 +51,7 @@ class Isolate;
 // DBG: Builtin in platform-dependent assembly, used by the debugger.
 //      Args: name
 
-#define BUILTIN_LIST(CPP, API, TFJ, TFS, ASM, ASH, DBG)                        \
+#define BUILTIN_LIST_BASE(CPP, API, TFJ, TFS, ASM, ASH, DBG)                   \
   ASM(Abort)                                                                   \
   /* Code aging */                                                             \
   CODE_AGE_LIST_WITH_ARG(DECLARE_CODE_AGE_BUILTIN, ASM)                        \
@@ -845,9 +845,9 @@ class Isolate;
   CPP(StringPrototypeToLocaleLowerCase)                                        \
   /* ES #sec-string.prototype.tolocaleuppercase */                             \
   CPP(StringPrototypeToLocaleUpperCase)                                        \
-  /* ES #sec-string.prototype.tolowercase */                                   \
+  /* (obsolete) Unibrow version */                                             \
   CPP(StringPrototypeToLowerCase)                                              \
-  /* ES #sec-string.prototype.touppercase */                                   \
+  /* (obsolete) Unibrow version */                                             \
   CPP(StringPrototypeToUpperCase)                                              \
   CPP(StringPrototypeTrim)                                                     \
   CPP(StringPrototypeTrimLeft)                                                 \
@@ -936,6 +936,19 @@ class Isolate;
   /* #sec-async-iterator-value-unwrap-functions */                             \
   TFJ(AsyncIteratorValueUnwrap, 1, kValue)
 
+#ifdef V8_I18N_SUPPORT
+#define BUILTIN_LIST(CPP, API, TFJ, TFS, ASM, ASH, DBG) \
+  BUILTIN_LIST_BASE(CPP, API, TFJ, TFS, ASM, ASH, DBG)  \
+                                                        \
+  /* ES #sec-string.prototype.tolowercase */            \
+  CPP(StringPrototypeToLowerCaseI18N)                   \
+  /* ES #sec-string.prototype.touppercase */            \
+  CPP(StringPrototypeToUpperCaseI18N)
+#else
+#define BUILTIN_LIST(CPP, API, TFJ, TFS, ASM, ASH, DBG) \
+  BUILTIN_LIST_BASE(CPP, API, TFJ, TFS, ASM, ASH, DBG)
+#endif  // V8_I18N_SUPPORT
+
 #define BUILTIN_PROMISE_REJECTION_PREDICTION_LIST(V) \
   V(AsyncFromSyncIteratorPrototypeNext)              \
   V(AsyncFromSyncIteratorPrototypeReturn)            \

src/i18n.cc

@@ -11,6 +11,7 @@
 #include "src/factory.h"
 #include "src/isolate.h"
 #include "src/objects-inl.h"
+#include "src/string-case.h"
 #include "unicode/brkiter.h"
 #include "unicode/calendar.h"
 #include "unicode/coll.h"
@@ -691,10 +692,8 @@ void SetResolvedBreakIteratorSettings(Isolate* isolate,
                           SLOPPY).Assert();
   }
 }
-
 }  // namespace
 
-
 // static
 icu::SimpleDateFormat* DateFormat::InitializeDateTimeFormat(
     Isolate* isolate,
@@ -909,5 +908,309 @@ void V8BreakIterator::DeleteBreakIterator(
   GlobalHandles::Destroy(reinterpret_cast<Object**>(data.GetParameter()));
 }
 
+namespace {
+inline bool IsASCIIUpper(uint16_t ch) { return ch >= 'A' && ch <= 'Z'; }
+
+const uint8_t kToLower[256] = {
+    0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0A, 0x0B,
+    0x0C, 0x0D, 0x0E, 0x0F, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
+    0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F, 0x20, 0x21, 0x22, 0x23,
+    0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2A, 0x2B, 0x2C, 0x2D, 0x2E, 0x2F,
+    0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3A, 0x3B,
+    0x3C, 0x3D, 0x3E, 0x3F, 0x40, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67,
+    0x68, 0x69, 0x6A, 0x6B, 0x6C, 0x6D, 0x6E, 0x6F, 0x70, 0x71, 0x72, 0x73,
+    0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7A, 0x5B, 0x5C, 0x5D, 0x5E, 0x5F,
+    0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6A, 0x6B,
+    0x6C, 0x6D, 0x6E, 0x6F, 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77,
+    0x78, 0x79, 0x7A, 0x7B, 0x7C, 0x7D, 0x7E, 0x7F, 0x80, 0x81, 0x82, 0x83,
+    0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0x8A, 0x8B, 0x8C, 0x8D, 0x8E, 0x8F,
+    0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0x9A, 0x9B,
+    0x9C, 0x9D, 0x9E, 0x9F, 0xA0, 0xA1, 0xA2, 0xA3, 0xA4, 0xA5, 0xA6, 0xA7,
+    0xA8, 0xA9, 0xAA, 0xAB, 0xAC, 0xAD, 0xAE, 0xAF, 0xB0, 0xB1, 0xB2, 0xB3,
+    0xB4, 0xB5, 0xB6, 0xB7, 0xB8, 0xB9, 0xBA, 0xBB, 0xBC, 0xBD, 0xBE, 0xBF,
+    0xE0, 0xE1, 0xE2, 0xE3, 0xE4, 0xE5, 0xE6, 0xE7, 0xE8, 0xE9, 0xEA, 0xEB,
+    0xEC, 0xED, 0xEE, 0xEF, 0xF0, 0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xD7,
+    0xF8, 0xF9, 0xFA, 0xFB, 0xFC, 0xFD, 0xFE, 0xDF, 0xE0, 0xE1, 0xE2, 0xE3,
+    0xE4, 0xE5, 0xE6, 0xE7, 0xE8, 0xE9, 0xEA, 0xEB, 0xEC, 0xED, 0xEE, 0xEF,
+    0xF0, 0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xF7, 0xF8, 0xF9, 0xFA, 0xFB,
+    0xFC, 0xFD, 0xFE, 0xFF,
+};
+
+inline uint16_t ToLatin1Lower(uint16_t ch) {
+  return static_cast<uint16_t>(kToLower[ch]);
+}
+
+inline uint16_t ToASCIIUpper(uint16_t ch) {
+  return ch & ~((ch >= 'a' && ch <= 'z') << 5);
+}
+
+// Does not work for U+00DF (sharp-s), U+00B5 (micron), U+00FF.
+inline uint16_t ToLatin1Upper(uint16_t ch) {
+  DCHECK(ch != 0xDF && ch != 0xB5 && ch != 0xFF);
+  return ch &
+         ~(((ch >= 'a' && ch <= 'z') || (((ch & 0xE0) == 0xE0) && ch != 0xF7))
+           << 5);
+}
+
+template <typename Char>
+bool ToUpperFastASCII(const Vector<const Char>& src,
+                      Handle<SeqOneByteString> result) {
+  // Do a faster loop for the case where all the characters are ASCII.
+  uint16_t ored = 0;
+  int32_t index = 0;
+  for (auto it = src.begin(); it != src.end(); ++it) {
+    uint16_t ch = static_cast<uint16_t>(*it);
+    ored |= ch;
+    result->SeqOneByteStringSet(index++, ToASCIIUpper(ch));
+  }
+  return !(ored & ~0x7F);
+}
+
+const uint16_t sharp_s = 0xDF;
+
+template <typename Char>
+bool ToUpperOneByte(const Vector<const Char>& src, uint8_t* dest,
+                    int* sharp_s_count) {
+  // Still pretty-fast path for the input with non-ASCII Latin-1 characters.
+
+  // There are two special cases.
+  //  1. U+00B5 and U+00FF are mapped to a character beyond U+00FF.
+  //  2. Lower case sharp-S converts to "SS" (two characters)
+  *sharp_s_count = 0;
+  for (auto it = src.begin(); it != src.end(); ++it) {
+    uint16_t ch = static_cast<uint16_t>(*it);
+    if (V8_UNLIKELY(ch == sharp_s)) {
+      ++(*sharp_s_count);
+      continue;
+    }
+    if (V8_UNLIKELY(ch == 0xB5 || ch == 0xFF)) {
+      // Since this upper-cased character does not fit in an 8-bit string, we
+      // need to take the 16-bit path.
+      return false;
+    }
+    *dest++ = ToLatin1Upper(ch);
+  }
+
+  return true;
+}
+
+template <typename Char>
+void ToUpperWithSharpS(const Vector<const Char>& src,
+                       Handle<SeqOneByteString> result) {
+  int32_t dest_index = 0;
+  for (auto it = src.begin(); it != src.end(); ++it) {
+    uint16_t ch = static_cast<uint16_t>(*it);
+    if (ch == sharp_s) {
+      result->SeqOneByteStringSet(dest_index++, 'S');
+      result->SeqOneByteStringSet(dest_index++, 'S');
+    } else {
+      result->SeqOneByteStringSet(dest_index++, ToLatin1Upper(ch));
+    }
+  }
+}
+
+inline int FindFirstUpperOrNonAscii(Handle<String> s, int length) {
+  for (int index = 0; index < length; ++index) {
+    uint16_t ch = s->Get(index);
+    if (V8_UNLIKELY(IsASCIIUpper(ch) || ch & ~0x7F)) {
+      return index;
+    }
+  }
+  return length;
+}
+
+}  // namespace
+
+const UChar* GetUCharBufferFromFlat(const String::FlatContent& flat,
+                                    std::unique_ptr<uc16[]>* dest,
+                                    int32_t length) {
+  DCHECK(flat.IsFlat());
+  if (flat.IsOneByte()) {
+    if (!*dest) {
+      dest->reset(NewArray<uc16>(length));
+      CopyChars(dest->get(), flat.ToOneByteVector().start(), length);
+    }
+    return reinterpret_cast<const UChar*>(dest->get());
+  } else {
+    return reinterpret_cast<const UChar*>(flat.ToUC16Vector().start());
+  }
+}
+
+MUST_USE_RESULT Object* LocaleConvertCase(Handle<String> s, Isolate* isolate,
+                                          bool is_to_upper, const char* lang) {
+  auto case_converter = is_to_upper ? u_strToUpper : u_strToLower;
+  int32_t src_length = s->length();
+  int32_t dest_length = src_length;
+  UErrorCode status;
+  Handle<SeqTwoByteString> result;
+  std::unique_ptr<uc16[]> sap;
+
+  if (dest_length == 0) return isolate->heap()->empty_string();
+
+  // This is not a real loop. It'll be executed only once (no overflow) or
+  // twice (overflow).
+  for (int i = 0; i < 2; ++i) {
+    // Case conversion can increase the string length (e.g. sharp-S => SS) so
+    // that we have to handle RangeError exceptions here.
+    ASSIGN_RETURN_FAILURE_ON_EXCEPTION(
+        isolate, result, isolate->factory()->NewRawTwoByteString(dest_length));
+    DisallowHeapAllocation no_gc;
+    DCHECK(s->IsFlat());
+    String::FlatContent flat = s->GetFlatContent();
+    const UChar* src = GetUCharBufferFromFlat(flat, &sap, src_length);
+    status = U_ZERO_ERROR;
+    dest_length = case_converter(reinterpret_cast<UChar*>(result->GetChars()),
+                                 dest_length, src, src_length, lang, &status);
+    if (status != U_BUFFER_OVERFLOW_ERROR) break;
+  }
+
+  // In most cases, the output will fill the destination buffer completely
+  // leading to an unterminated string (U_STRING_NOT_TERMINATED_WARNING).
+  // Only in rare cases, it'll be shorter than the destination buffer and
+  // |result| has to be truncated.
+  DCHECK(U_SUCCESS(status));
+  if (V8_LIKELY(status == U_STRING_NOT_TERMINATED_WARNING)) {
+    DCHECK(dest_length == result->length());
+    return *result;
+  }
+  if (U_SUCCESS(status)) {
+    DCHECK(dest_length < result->length());
+    return *Handle<SeqTwoByteString>::cast(
+        SeqString::Truncate(result, dest_length));
+  }
+  return *s;
+}
+
+MUST_USE_RESULT Object* ConvertToLower(Handle<String> s, Isolate* isolate) {
+  if (!s->HasOnlyOneByteChars()) {
+    // Use a slower implementation for strings with characters beyond U+00FF.
+    return LocaleConvertCase(s, isolate, false, "");
+  }
+
+  int length = s->length();
+
+  // We depend here on the invariant that the length of a Latin1
+  // string is invariant under ToLowerCase, and the result always
+  // fits in the Latin1 range in the *root locale*. It does not hold
+  // for ToUpperCase even in the root locale.
+
+  // Scan the string for uppercase and non-ASCII characters for strings
+  // shorter than a machine-word without any memory allocation overhead.
+  // TODO(jshin): Apply this to a longer input by breaking FastAsciiConvert()
+  // to two parts, one for scanning the prefix with no change and the other for
+  // handling ASCII-only characters.
+  int index_to_first_unprocessed = length;
+  const bool is_short = length < static_cast<int>(sizeof(uintptr_t));
+  if (is_short) {
+    index_to_first_unprocessed = FindFirstUpperOrNonAscii(s, length);
+    // Nothing to do if the string is all ASCII with no uppercase.
+    if (index_to_first_unprocessed == length) return *s;
+  }
+
+  Handle<SeqOneByteString> result =
+      isolate->factory()->NewRawOneByteString(length).ToHandleChecked();
+
+  DisallowHeapAllocation no_gc;
+  DCHECK(s->IsFlat());
+  String::FlatContent flat = s->GetFlatContent();
+  uint8_t* dest = result->GetChars();
+  if (flat.IsOneByte()) {
+    const uint8_t* src = flat.ToOneByteVector().start();
+    bool has_changed_character = false;
+    index_to_first_unprocessed = FastAsciiConvert<true>(
+        reinterpret_cast<char*>(dest), reinterpret_cast<const char*>(src),
+        length, &has_changed_character);
+    // If not ASCII, we keep the result up to index_to_first_unprocessed and
+    // process the rest.
+    if (index_to_first_unprocessed == length)
+      return has_changed_character ? *result : *s;
+
+    for (int index = index_to_first_unprocessed; index < length; ++index) {
+      dest[index] = ToLatin1Lower(static_cast<uint16_t>(src[index]));
+    }
+  } else {
+    if (index_to_first_unprocessed == length) {
+      DCHECK(!is_short);
+      index_to_first_unprocessed = FindFirstUpperOrNonAscii(s, length);
+    }
+    // Nothing to do if the string is all ASCII with no uppercase.
+    if (index_to_first_unprocessed == length) return *s;
+    const uint16_t* src = flat.ToUC16Vector().start();
+    CopyChars(dest, src, index_to_first_unprocessed);
+    for (int index = index_to_first_unprocessed; index < length; ++index) {
+      dest[index] = ToLatin1Lower(static_cast<uint16_t>(src[index]));
+    }
+  }
+
+  return *result;
+}
+
+MUST_USE_RESULT Object* ConvertToUpper(Handle<String> s, Isolate* isolate) {
+  int32_t length = s->length();
+  if (s->HasOnlyOneByteChars() && length > 0) {
+    Handle<SeqOneByteString> result =
+        isolate->factory()->NewRawOneByteString(length).ToHandleChecked();
+
+    DCHECK(s->IsFlat());
+    int sharp_s_count;
+    bool is_result_single_byte;
+    {
+      DisallowHeapAllocation no_gc;
+      String::FlatContent flat = s->GetFlatContent();
+      uint8_t* dest = result->GetChars();
+      if (flat.IsOneByte()) {
+        Vector<const uint8_t> src = flat.ToOneByteVector();
+        bool has_changed_character = false;
+        int index_to_first_unprocessed =
+            FastAsciiConvert<false>(reinterpret_cast<char*>(result->GetChars()),
+                                    reinterpret_cast<const char*>(src.start()),
+                                    length, &has_changed_character);
+        if (index_to_first_unprocessed == length)
+          return has_changed_character ? *result : *s;
+        // If not ASCII, we keep the result up to index_to_first_unprocessed and
+        // process the rest.
+        is_result_single_byte =
+            ToUpperOneByte(src.SubVector(index_to_first_unprocessed, length),
+                           dest + index_to_first_unprocessed, &sharp_s_count);
+      } else {
+        DCHECK(flat.IsTwoByte());
+        Vector<const uint16_t> src = flat.ToUC16Vector();
+        if (ToUpperFastASCII(src, result)) return *result;
+        is_result_single_byte = ToUpperOneByte(src, dest, &sharp_s_count);
+      }
+    }
+
+    // Go to the full Unicode path if there are characters whose uppercase
+    // is beyond the Latin-1 range (cannot be represented in OneByteString).
+    if (V8_UNLIKELY(!is_result_single_byte)) {
+      return LocaleConvertCase(s, isolate, true, "");
+    }
+
+    if (sharp_s_count == 0) return *result;
+
+    // We have sharp_s_count sharp-s characters, but the result is still
+    // in the Latin-1 range.
+    ASSIGN_RETURN_FAILURE_ON_EXCEPTION(
+        isolate, result,
+        isolate->factory()->NewRawOneByteString(length + sharp_s_count));
+    DisallowHeapAllocation no_gc;
+    String::FlatContent flat = s->GetFlatContent();
+    if (flat.IsOneByte()) {
+      ToUpperWithSharpS(flat.ToOneByteVector(), result);
+    } else {
+      ToUpperWithSharpS(flat.ToUC16Vector(), result);
+    }
+
+    return *result;
+  }
+
+  return LocaleConvertCase(s, isolate, true, "");
+}
+
+MUST_USE_RESULT Object* ConvertCase(Handle<String> s, bool is_upper,
+                                    Isolate* isolate) {
+  return is_upper ? ConvertToUpper(s, isolate) : ConvertToLower(s, isolate);
+}
+
 }  // namespace internal
 }  // namespace v8

src/i18n.h

@@ -128,6 +128,16 @@ class V8BreakIterator {
   V8BreakIterator();
 };
 
+const UChar* GetUCharBufferFromFlat(const String::FlatContent& flat,
+                                    std::unique_ptr<uc16[]>* dest,
+                                    int32_t length);
+MUST_USE_RESULT Object* LocaleConvertCase(Handle<String> s, Isolate* isolate,
+                                          bool is_to_upper, const char* lang);
+MUST_USE_RESULT Object* ConvertToLower(Handle<String> s, Isolate* isolate);
+MUST_USE_RESULT Object* ConvertToUpper(Handle<String> s, Isolate* isolate);
+MUST_USE_RESULT Object* ConvertCase(Handle<String> s, bool is_upper,
+                                    Isolate* isolate);
+
 }  // namespace internal
 }  // namespace v8
 

src/runtime/runtime-i18n.cc

@@ -15,7 +15,6 @@
 #include "src/i18n.h"
 #include "src/isolate-inl.h"
 #include "src/messages.h"
-#include "src/string-case.h"
 #include "src/utils.h"
 
 #include "unicode/brkiter.h"
@@ -49,24 +48,6 @@
 
 namespace v8 {
 namespace internal {
-namespace {
-
-const UChar* GetUCharBufferFromFlat(const String::FlatContent& flat,
-                                    std::unique_ptr<uc16[]>* dest,
-                                    int32_t length) {
-  DCHECK(flat.IsFlat());
-  if (flat.IsOneByte()) {
-    if (!*dest) {
-      dest->reset(NewArray<uc16>(length));
-      CopyChars(dest->get(), flat.ToOneByteVector().start(), length);
-    }
-    return reinterpret_cast<const UChar*>(dest->get());
-  } else {
-    return reinterpret_cast<const UChar*>(flat.ToUC16Vector().start());
-  }
-}
-
-}  // namespace
 
 // ECMA 402 6.2.3
 RUNTIME_FUNCTION(Runtime_CanonicalizeLanguageTag) {
@@ -823,295 +804,6 @@ RUNTIME_FUNCTION(Runtime_BreakIteratorBreakType) {
   }
 }
 
-namespace {
-MUST_USE_RESULT Object* LocaleConvertCase(Handle<String> s, Isolate* isolate,
-                                          bool is_to_upper, const char* lang) {
-  auto case_converter = is_to_upper ? u_strToUpper : u_strToLower;
-  int32_t src_length = s->length();
-  int32_t dest_length = src_length;
-  UErrorCode status;
-  Handle<SeqTwoByteString> result;
-  std::unique_ptr<uc16[]> sap;
-
-  if (dest_length == 0) return isolate->heap()->empty_string();
-
-  // This is not a real loop. It'll be executed only once (no overflow) or
-  // twice (overflow).
-  for (int i = 0; i < 2; ++i) {
-    // Case conversion can increase the string length (e.g. sharp-S => SS) so
-    // that we have to handle RangeError exceptions here.
-    ASSIGN_RETURN_FAILURE_ON_EXCEPTION(
-        isolate, result, isolate->factory()->NewRawTwoByteString(dest_length));
-    DisallowHeapAllocation no_gc;
-    DCHECK(s->IsFlat());
-    String::FlatContent flat = s->GetFlatContent();
-    const UChar* src = GetUCharBufferFromFlat(flat, &sap, src_length);
-    status = U_ZERO_ERROR;
-    dest_length = case_converter(reinterpret_cast<UChar*>(result->GetChars()),
-                                 dest_length, src, src_length, lang, &status);
-    if (status != U_BUFFER_OVERFLOW_ERROR) break;
-  }
-
-  // In most cases, the output will fill the destination buffer completely
-  // leading to an unterminated string (U_STRING_NOT_TERMINATED_WARNING).
-  // Only in rare cases, it'll be shorter than the destination buffer and
-  // |result| has to be truncated.
-  DCHECK(U_SUCCESS(status));
-  if (V8_LIKELY(status == U_STRING_NOT_TERMINATED_WARNING)) {
-    DCHECK(dest_length == result->length());
-    return *result;
-  }
-  if (U_SUCCESS(status)) {
-    DCHECK(dest_length < result->length());
-    return *Handle<SeqTwoByteString>::cast(
-        SeqString::Truncate(result, dest_length));
-  }
-  return *s;
-}
-
-inline bool IsASCIIUpper(uint16_t ch) { return ch >= 'A' && ch <= 'Z'; }
-
-const uint8_t kToLower[256] = {
-    0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0A, 0x0B,
-    0x0C, 0x0D, 0x0E, 0x0F, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
-    0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F, 0x20, 0x21, 0x22, 0x23,
-    0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2A, 0x2B, 0x2C, 0x2D, 0x2E, 0x2F,
-    0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3A, 0x3B,
-    0x3C, 0x3D, 0x3E, 0x3F, 0x40, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67,
-    0x68, 0x69, 0x6A, 0x6B, 0x6C, 0x6D, 0x6E, 0x6F, 0x70, 0x71, 0x72, 0x73,
-    0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7A, 0x5B, 0x5C, 0x5D, 0x5E, 0x5F,
-    0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6A, 0x6B,
-    0x6C, 0x6D, 0x6E, 0x6F, 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77,
-    0x78, 0x79, 0x7A, 0x7B, 0x7C, 0x7D, 0x7E, 0x7F, 0x80, 0x81, 0x82, 0x83,
-    0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0x8A, 0x8B, 0x8C, 0x8D, 0x8E, 0x8F,
-    0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0x9A, 0x9B,
-    0x9C, 0x9D, 0x9E, 0x9F, 0xA0, 0xA1, 0xA2, 0xA3, 0xA4, 0xA5, 0xA6, 0xA7,
-    0xA8, 0xA9, 0xAA, 0xAB, 0xAC, 0xAD, 0xAE, 0xAF, 0xB0, 0xB1, 0xB2, 0xB3,
-    0xB4, 0xB5, 0xB6, 0xB7, 0xB8, 0xB9, 0xBA, 0xBB, 0xBC, 0xBD, 0xBE, 0xBF,
-    0xE0, 0xE1, 0xE2, 0xE3, 0xE4, 0xE5, 0xE6, 0xE7, 0xE8, 0xE9, 0xEA, 0xEB,
-    0xEC, 0xED, 0xEE, 0xEF, 0xF0, 0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xD7,
-    0xF8, 0xF9, 0xFA, 0xFB, 0xFC, 0xFD, 0xFE, 0xDF, 0xE0, 0xE1, 0xE2, 0xE3,
-    0xE4, 0xE5, 0xE6, 0xE7, 0xE8, 0xE9, 0xEA, 0xEB, 0xEC, 0xED, 0xEE, 0xEF,
-    0xF0, 0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xF7, 0xF8, 0xF9, 0xFA, 0xFB,
-    0xFC, 0xFD, 0xFE, 0xFF,
-};
-
-inline uint16_t ToLatin1Lower(uint16_t ch) {
-  return static_cast<uint16_t>(kToLower[ch]);
-}
-
-inline uint16_t ToASCIIUpper(uint16_t ch) {
-  return ch & ~((ch >= 'a' && ch <= 'z') << 5);
-}
-
-// Does not work for U+00DF (sharp-s), U+00B5 (micron), U+00FF.
-inline uint16_t ToLatin1Upper(uint16_t ch) {
-  DCHECK(ch != 0xDF && ch != 0xB5 && ch != 0xFF);
-  return ch &
-         ~(((ch >= 'a' && ch <= 'z') || (((ch & 0xE0) == 0xE0) && ch != 0xF7))
-           << 5);
-}
-
-template <typename Char>
-bool ToUpperFastASCII(const Vector<const Char>& src,
-                      Handle<SeqOneByteString> result) {
-  // Do a faster loop for the case where all the characters are ASCII.
-  uint16_t ored = 0;
-  int32_t index = 0;
-  for (auto it = src.begin(); it != src.end(); ++it) {
-    uint16_t ch = static_cast<uint16_t>(*it);
-    ored |= ch;
-    result->SeqOneByteStringSet(index++, ToASCIIUpper(ch));
-  }
-  return !(ored & ~0x7F);
-}
-
-const uint16_t sharp_s = 0xDF;
-
-template <typename Char>
-bool ToUpperOneByte(const Vector<const Char>& src, uint8_t* dest,
-                    int* sharp_s_count) {
-  // Still pretty-fast path for the input with non-ASCII Latin-1 characters.
-
-  // There are two special cases.
-  //  1. U+00B5 and U+00FF are mapped to a character beyond U+00FF.
-  //  2. Lower case sharp-S converts to "SS" (two characters)
-  *sharp_s_count = 0;
-  for (auto it = src.begin(); it != src.end(); ++it) {
-    uint16_t ch = static_cast<uint16_t>(*it);
-    if (V8_UNLIKELY(ch == sharp_s)) {
-      ++(*sharp_s_count);
-      continue;
-    }
-    if (V8_UNLIKELY(ch == 0xB5 || ch == 0xFF)) {
-      // Since this upper-cased character does not fit in an 8-bit string, we
-      // need to take the 16-bit path.
-      return false;
-    }
-    *dest++ = ToLatin1Upper(ch);
-  }
-
-  return true;
-}
-
-template <typename Char>
-void ToUpperWithSharpS(const Vector<const Char>& src,
-                       Handle<SeqOneByteString> result) {
-  int32_t dest_index = 0;
-  for (auto it = src.begin(); it != src.end(); ++it) {
-    uint16_t ch = static_cast<uint16_t>(*it);
-    if (ch == sharp_s) {
-      result->SeqOneByteStringSet(dest_index++, 'S');
-      result->SeqOneByteStringSet(dest_index++, 'S');
-    } else {
-      result->SeqOneByteStringSet(dest_index++, ToLatin1Upper(ch));
-    }
-  }
-}
-
-inline int FindFirstUpperOrNonAscii(Handle<String> s, int length) {
-  for (int index = 0; index < length; ++index) {
-    uint16_t ch = s->Get(index);
-    if (V8_UNLIKELY(IsASCIIUpper(ch) || ch & ~0x7F)) {
-      return index;
-    }
-  }
-  return length;
-}
-
-MUST_USE_RESULT Object* ConvertToLower(Handle<String> s, Isolate* isolate) {
-  if (!s->HasOnlyOneByteChars()) {
-    // Use a slower implementation for strings with characters beyond U+00FF.
-    return LocaleConvertCase(s, isolate, false, "");
-  }
-
-  int length = s->length();
-
-  // We depend here on the invariant that the length of a Latin1
-  // string is invariant under ToLowerCase, and the result always
-  // fits in the Latin1 range in the *root locale*. It does not hold
-  // for ToUpperCase even in the root locale.
-
-  // Scan the string for uppercase and non-ASCII characters for strings
-  // shorter than a machine-word without any memory allocation overhead.
-  // TODO(jshin): Apply this to a longer input by breaking FastAsciiConvert()
-  // to two parts, one for scanning the prefix with no change and the other for
-  // handling ASCII-only characters.
-  int index_to_first_unprocessed = length;
-  const bool is_short = length < static_cast<int>(sizeof(uintptr_t));
-  if (is_short) {
-    index_to_first_unprocessed = FindFirstUpperOrNonAscii(s, length);
-    // Nothing to do if the string is all ASCII with no uppercase.
-    if (index_to_first_unprocessed == length) return *s;
-  }
-
-  Handle<SeqOneByteString> result =
-      isolate->factory()->NewRawOneByteString(length).ToHandleChecked();
-
-  DisallowHeapAllocation no_gc;
-  DCHECK(s->IsFlat());
-  String::FlatContent flat = s->GetFlatContent();
-  uint8_t* dest = result->GetChars();
-  if (flat.IsOneByte()) {
-    const uint8_t* src = flat.ToOneByteVector().start();
-    bool has_changed_character = false;
-    index_to_first_unprocessed = FastAsciiConvert<true>(
-        reinterpret_cast<char*>(dest), reinterpret_cast<const char*>(src),
-        length, &has_changed_character);
-    // If not ASCII, we keep the result up to index_to_first_unprocessed and
-    // process the rest.
-    if (index_to_first_unprocessed == length)
-      return has_changed_character ? *result : *s;
-
-    for (int index = index_to_first_unprocessed; index < length; ++index) {
-      dest[index] = ToLatin1Lower(static_cast<uint16_t>(src[index]));
-    }
-  } else {
-    if (index_to_first_unprocessed == length) {
-      DCHECK(!is_short);
-      index_to_first_unprocessed = FindFirstUpperOrNonAscii(s, length);
-    }
-    // Nothing to do if the string is all ASCII with no uppercase.
-    if (index_to_first_unprocessed == length) return *s;
-    const uint16_t* src = flat.ToUC16Vector().start();
-    CopyChars(dest, src, index_to_first_unprocessed);
-    for (int index = index_to_first_unprocessed; index < length; ++index) {
-      dest[index] = ToLatin1Lower(static_cast<uint16_t>(src[index]));
-    }
-  }
-
-  return *result;
-}
-
-MUST_USE_RESULT Object* ConvertToUpper(Handle<String> s, Isolate* isolate) {
-  int32_t length = s->length();
-  if (s->HasOnlyOneByteChars() && length > 0) {
-    Handle<SeqOneByteString> result =
-        isolate->factory()->NewRawOneByteString(length).ToHandleChecked();
-
-    DCHECK(s->IsFlat());
-    int sharp_s_count;
-    bool is_result_single_byte;
-    {
-      DisallowHeapAllocation no_gc;
-      String::FlatContent flat = s->GetFlatContent();
-      uint8_t* dest = result->GetChars();
-      if (flat.IsOneByte()) {
-        Vector<const uint8_t> src = flat.ToOneByteVector();
-        bool has_changed_character = false;
-        int index_to_first_unprocessed =
-            FastAsciiConvert<false>(reinterpret_cast<char*>(result->GetChars()),
-                                    reinterpret_cast<const char*>(src.start()),
-                                    length, &has_changed_character);
-        if (index_to_first_unprocessed == length)
-          return has_changed_character ? *result : *s;
-        // If not ASCII, we keep the result up to index_to_first_unprocessed and
-        // process the rest.
-        is_result_single_byte =
-            ToUpperOneByte(src.SubVector(index_to_first_unprocessed, length),
-                           dest + index_to_first_unprocessed, &sharp_s_count);
-      } else {
-        DCHECK(flat.IsTwoByte());
-        Vector<const uint16_t> src = flat.ToUC16Vector();
-        if (ToUpperFastASCII(src, result)) return *result;
-        is_result_single_byte = ToUpperOneByte(src, dest, &sharp_s_count);
-      }
-    }
-
-    // Go to the full Unicode path if there are characters whose uppercase
-    // is beyond the Latin-1 range (cannot be represented in OneByteString).
-    if (V8_UNLIKELY(!is_result_single_byte)) {
-      return LocaleConvertCase(s, isolate, true, "");
-    }
-
-    if (sharp_s_count == 0) return *result;
-
-    // We have sharp_s_count sharp-s characters, but the result is still
-    // in the Latin-1 range.
-    ASSIGN_RETURN_FAILURE_ON_EXCEPTION(
-        isolate, result,
-        isolate->factory()->NewRawOneByteString(length + sharp_s_count));
-    DisallowHeapAllocation no_gc;
-    String::FlatContent flat = s->GetFlatContent();
-    if (flat.IsOneByte()) {
-      ToUpperWithSharpS(flat.ToOneByteVector(), result);
-    } else {
-      ToUpperWithSharpS(flat.ToUC16Vector(), result);
-    }
-
-    return *result;
-  }
-
-  return LocaleConvertCase(s, isolate, true, "");
-}
-
-MUST_USE_RESULT Object* ConvertCase(Handle<String> s, bool is_upper,
-                                    Isolate* isolate) {
-  return is_upper ? ConvertToUpper(s, isolate) : ConvertToLower(s, isolate);
-}
-
-}  // namespace
-
 RUNTIME_FUNCTION(Runtime_StringToLowerCaseI18N) {
   HandleScope scope(isolate);
   DCHECK_EQ(args.length(), 1);

src/v8.gyp

@@ -527,6 +527,7 @@
         'builtins/builtins-sharedarraybuffer-gen.cc',
         'builtins/builtins-string.cc',
         'builtins/builtins-string-gen.cc',
+        'builtins/builtins-intl.cc',
         'builtins/builtins-symbol.cc',
         'builtins/builtins-symbol-gen.cc',
         'builtins/builtins-typedarray.cc',